My forecast data from the VRM dashboard (under Installation data > Overview) currently appears to be implausible, showing roughly double the correct figures, for both Solar production and Consumption.
The data displayed in the graphs reflects that in the totals. I also get the same figure for Solar production forecast from an API call (~55 kWh) - (this shows a doubling for both the MPPT and the FoxESS inverter on AC Out)
Clicking forward in time on VRM yields similar results, until it gets to the 11th August, when the Solar appears feasible again, with the Consumption following suit from the 12th onwards.
All historical data from the CerboGX appears normal (I log everything to influxdb).
Anyone else seeing similar, or is it just me? I’ve been using the VRM forecasts for setting my overnight charge targets (because they’ve been so accurate and reliable so far), hence my noticing.
Thanks.
[Cerbo GX, MPPT 250/100, Pylontech US3000Cs, single phase MPii, FoxESS on AC out with EM112 over RS485].
For me, it also is almost doubled. For my installation, pv ertrag Postleitzahl shows a span of 3.84 to 5.09 kWh/kWp, which roughly corresponds to 67.8 to 89,6 kWh yield. VRM predicts a range of 147kWh to 179kWh. wetteronline.de forecast is 11h of sunshine with some clouds in the morning. I don’t use the forecast for any automation, though, and don’t charge from grid.
For me the solar forecasts in API call is 0 but on VRM its doubled and is so until the 11th as you other noted also.
The consumption forecast seems doubled aswell.
Looking at my Android phone and on the VRM widget that I have setup to show the status of the system it just flashes like it tries to refresh/load every second without showing any data.
@Barbara Something that happened in the VRM backend over night since yesterday everything was normal in the predictions.
Hi Nick, good to hear the team is on it. (Same issue here)
Is the ‘issue’ of yesterday’s VE-EVCS consumption ‘lingering’ on every forecasted hour (for days to come) also on the list? One would expect that a very known consumption would/could lead to a pretty defined forecast. Currently those kWh seem to both ‘load’ the ‘total day expected average’ AND ‘on the hour expected’ leading to doubling within the days ahead.
Many thanks, Nick & whoever is working on the issue… also thanks to all who replied. Good to know I’m not going mad, also that I still haven’t quite found the limits as far as Node-Red abuse goes.
We’ve pushed a change, that should improve the situation. Please let me know if you’re still seeing issues!
As you might’ve experienced, we’re having some issues with our timeseries database. This resulted in multiple values being returned where we expected a single value. These were summed, resulting in the doubled or tripled data.
Sorry for the inconvenience, hope all is well now!
While today’s forecast seems to have somewhat corrected itself from the ‘double-ish’ figures to more reasonable values, tomorrow’s forecast (and the days thereafter) is now way too low (about 30-50% of nominal yields on sunny days) if the view is selected on a daily basis (tomorrow, day after, …). If the 7-day view is selected, the figures are still 200% (same as this morning) for the entire week!
I can see the same ‘7 day future prediction’ is still high. Normal max PV yield 25-29kWh and consumption 12-15 + EVCS (10/20/30 intermittent, 2nd-3rd-4th day)
Maybe long term still has to recalculate with the adaptation @jarco mentioned?
To echo Stanislav above, I’m now seeing (seemingly) correct data on the VRM portal, but the data returned from API calls is completely different, now roughly showing triple (3x) the expected yields, e.g.:
Anyway, thanks for your efforts - it’s encouraging to know that the ‘good’ data is still in there somewhere, even if it’s not currently being spat out by the API. Good luck!
As of now (05:45 AM local time), the forecast in VRM for my location for today is way too low (predicts 45 to 55, should be in the range of 100 to 130)
